Website personalization and interactive assistant

Abstract

Methods, systems, and computer programs are presented to provide a personalized view of webpages and an interactive assistant for exploring website content. One method includes operations for accessing website data of a website, and detecting a request entered by a user in a user interface (UI). One or more blocks for a webpage are generated in response to the request, where generating each block includes constructing a prompt by entering website data, user information, and block information on a prompt template; providing the prompt as input to a Generative Artificial Intelligence (GAI) tool; and utilizing the response from the GAI tool to build the block for the webpage. The method further includes causing presentation on the UI of the webpage including the generated one or more blocks.

What is provisionally claimed is: 
     
       1. A computer-implemented method comprising:
 accessing website data of a website; 
 detecting a request entered by a user in a user interface (UI) for a webpage in the website; 
 generating one or more blocks for the webpage in response to the request, wherein generating each block comprises:
 constructing a prompt based on the website data, user information, block information, and a prompt template, the prompt comprising descriptive text for desired results, the prompt template including fillable fields, wherein constructing the prompt comprises filling the fillable fields with information based on the website data and the user information; 
 providing the prompt, comprising website data, user information, and block information, as input to a Generative Artificial Intelligence (GAI) tool; 
 utilizing the response from the GAI tool to build the block for the webpage; and 
 adding the block built by the GAI tool to the webpage; and 
 
 causing presentation on the UI of the webpage including the generated one or more blocks. 
 
     
     
       2. The method as recited in  claim 1 , wherein accessing the website data further comprises:
 obtaining product information from the website; 
 identifying blocks of pages in the website; and 
 retrieving one or more entity tags for each block, wherein the website data comprises the product information, the identified blocks of pages in the website, and the one or more entity tags for each block. 
 
     
     
       3. The method as recited in  claim 1 , wherein accessing the website data further comprises:
 generating semantic embeddings for blocks in the webpages of the website. 
 
     
     
       4. The method as recited in  claim 3 , wherein the block is an image block for an image, wherein generating the semantic embeddings for the image block comprises:
 extracting captions from the image; 
 extracting a description of the image; 
 obtaining alt-text for the image; 
 obtaining one or more bounding boxes for the image; and 
 generating the semantic embedding for the image block based on the captions, the description, the alt-text, and the one or more bounding boxes. 
 
     
     
       5. The method as recited in  claim 3 , wherein generating the one or more blocks further comprises:
 generating a request embedding for the request; and 
 selecting the one or more blocks with the semantic embeddings closest to the request embedding. 
 
     
     
       6. The method as recited in  claim 1 , further comprising:
 providing an interactive assistant on the webpage in the UI, wherein the request is entered in the interactive assistant. 
 
     
     
       7. The method as recited in  claim 6 , further comprising:
 creating a prompt based on the request and the user information; 
 providing the prompt as input to a Large Language Model (LLM); and 
 presenting the response from the LLM in the interactive assistant. 
 
     
     
       8. The method as recited in  claim 1 , wherein the one or more blocks comprise an image block, wherein the GAI tool is an image-generation tool. 
     
     
       9. The method as recited in  claim 1 , wherein the one or more blocks comprise at least one image block and at least one text block. 
     
     
       10. The method as recited in  claim 1 , further comprising:
 after accessing the website data, performing incremental training for a Large Language Model (LLM) with text data from the website data. 
 
     
     
       11. A system comprising:
 a memory comprising instructions; and 
 one or more computer processors, wherein the instructions, when executed by the one or more computer processors, cause the system to perform operations comprising:
 accessing website data of a website; 
 detecting a request entered by a user in a user interface (UI) for a webpage in the website; 
 generating one or more blocks for the webpage in response to the request, wherein generating each block comprises:
 constructing a prompt based on the website data, user information, block information, and a prompt template, the prompt comprising descriptive text for desired results, the prompt template including fillable fields, wherein constructing the prompt comprises filling the fillable fields with information based on the website data and the user information; 
 providing the prompt, comprising website data, user information, and block information, as input to a Generative Artificial Intelligence (GAI) tool; 
 utilizing the response from the GAI tool to build the block for the webpage; and 
 adding the block built by the GAI tool to the webpage; and 
 
 causing presentation on the UI of the webpage including the generated one or more blocks. 
 
 
     
     
       12. The system as recited in  claim 11 , wherein accessing the website data further comprises:
 obtaining product information from the website; 
 identifying blocks of pages in the website; and 
 retrieving one or more entity tags for each block, wherein the website data comprises the product information, the identified blocks of pages in the website, and the one or more entity tags for each block. 
 
     
     
       13. The system as recited in  claim 11 , wherein accessing the website data further comprises:
 generating semantic embeddings for blocks in the webpages of the website. 
 
     
     
       14. The system as recited in  claim 13 , wherein the block is an image block for an image, wherein generating the semantic embeddings for the image block comprises:
 extracting captions from the image; 
 extracting a description of the image; 
 obtaining alt-text for the image; 
 obtaining one or more bounding boxes for the image; and 
 generating the semantic embedding for the image block based on the captions, the description, the alt-text, and the one or more bounding boxes. 
 
     
     
       15. The system as recited in  claim 13 , wherein generating the one or more blocks further comprises:
 generating a request embedding for the request; and 
 selecting the one or more blocks with the semantic embeddings closest to the request embedding. 
 
     
     
       16. A non-transitory machine-readable storage medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:
 accessing website data of a website; 
 detecting a request entered by a user in a user interface (UI) for a webpage in the website; 
 generating one or more blocks for the webpage in response to the request, wherein generating each block comprises:
 constructing a prompt based on the website data, user information, block information, and a prompt template, the prompt comprising descriptive text for desired results, the prompt template including fillable fields, wherein constructing the prompt comprises filling the fillable fields with information based on the website data and the user information; 
 providing the prompt, comprising website data, user information, and block information, as input to a Generative Artificial Intelligence (GAI) tool; 
 utilizing the response from the GAI tool to build the block for the webpage; and 
 adding the block built by the GAI tool to the webpage; and 
 
 causing presentation on the UI of the webpage including the generated one or more blocks. 
 
     
     
       17. The non-transitory machine-readable storage medium as recited in  claim 16 , wherein accessing the website data further comprises:
 obtaining product information from the website; 
 identifying blocks of pages in the website; and 
 retrieving one or more entity tags for each block, wherein the website data comprises the product information, the identified blocks of pages in the website, and the one or more entity tags for each block. 
 
     
     
       18. The non-transitory machine-readable storage medium as recited in  claim 16 , wherein accessing the website data further comprises:
 generating semantic embeddings for blocks in the webpages of the website. 
 
     
     
       19. The non-transitory machine-readable storage medium as recited in  claim 18 , wherein the block is an image block for an image, wherein generating the semantic embeddings for the image block comprises:
 extracting captions from the image; 
 extracting a description of the image; 
 obtaining alt-text for the image; 
 obtaining one or more bounding boxes for the image; and 
 generating the semantic embedding for the image block based on the captions, the description, the alt-text, and the one or more bounding boxes. 
 
     
     
       20. The non-transitory machine-readable storage medium as recited in  claim 18 , wherein generating the one or more blocks further comprises:
 generating a request embedding for the request, and 
 selecting the one or more blocks with the semantic embeddings closest to the request embedding.